They Were About to Hire for a Job They Could Not Yet Define

Context

The company sells practice management software to multi-site veterinary groups. Average contract value sits around $31,000, deals take a bit over two months, and they had just raised a Series A on the back of a founder and two account executives who knew every customer by name.

The round bought a bigger team. Five AEs, two SDRs, a first real marketing hire, a CS lead and two CS managers, a VP of Sales. Fourteen people in go-to-market where there had been five.

What the round did not buy was anyone whose job was the system all fourteen depended on.

By the time we met them the stack had grown to nine tools. Three had been bought by whoever felt the pain first and expensed it. The CRM had been configured in the company's first month, by a founder, at night, and never revisited.

The problem

Nobody was doing a bad job. That was the confusing part. Every function worked. The joins between them did not.

Lead routing lived in a spreadsheet called Routing_FINAL_v4, which one person updated by hand every Monday morning. Median time from form fill to first human contact was about 31 hours. The published cross-industry average is worse, around 42 hours, which is the kind of fact that makes a team feel fine about a number they should not feel fine about. During one week in March that person was on vacation and 60-odd leads sat untouched until Thursday.

Opportunity stages were still the CRM defaults from installation day. They described a sales process the company had stopped running two years earlier, so reps worked around them, which meant stage data described nothing at all.

Marketing reported about 340 qualified leads a month. Sales privately thought the real number was closer to 40. Both were arguing from their own dashboard and neither had ever written the definition down.

Just under 150 opportunities sat open with close dates already in the past. The oldest was 19 months old. Nobody wanted to mark them lost, because marking them lost felt like admitting something, so the reported win rate of 34% was measuring a pipeline that had quietly stopped resolving.

Three people could pull a pipeline report and produce three different answers. So the founder stopped asking for the report and read the deal list directly, which works at 20 deals and fails at 200.

Then the board asked for a hiring plan and the default answer was a Director of RevOps at $165K base, roughly $205K fully loaded. The trouble with that answer was that nobody in the company could describe what the role would do on day 30, because nobody had mapped what was actually broken. They were about to spend a fifth of a million dollars a year, plus a quarter of ramp, on a guess.

What we did

We embedded as their RevOps function rather than advising from beside it. Their standups, our own CRM logins, a backlog the whole team could see.

Weeks 1 to 2, measure before touching anything

Two weeks of watching how work actually moved produced a little over thirty distinct failures, each with a cost attached. Nine of them accounted for most of the pain.

Weeks 3 to 6, routing and the lead definition

Routing came off the spreadsheet and into rules the CRM enforces, with a 15-minute response SLA on inbound. On breach the lead reassigns round-robin to the next available rep rather than escalating to a manager, because at five AEs the manager is in a meeting and the lead is the thing that cannot wait. The 60-odd leads stranded during that March vacation week were pulled back into the queue and worked. Four were still live. One closed in August.

Territory came next, because it had to. Routing rules need somewhere to route to, and the account book had never been formally divided. Two AEs had spent the better part of a year quietly working the same regional chains. We split the book by group size and geography, wrote rules of engagement for the overlaps, and set quota against the new lines.

Then marketing and sales sat in one room and agreed a single definition of a qualified lead. Marketing had been reporting 340 a month. Sales had been muttering that the real number was about 40. Applying the agreed definition honestly to a month of history, the answer came out at 44.

Sales had been right for a year, and nobody had settled it because settling it required someone to own the definition rather than win the argument.

That looked like an 87% cut to marketing and it was not. It was the same demand, minus the volume nobody had been working. Sales now worked all 44 of them inside the SLA, and pipeline created rose 24% in the first full month after the change.

We also fixed the thing sitting underneath all of it. Lead source was being overwritten the moment a lead converted to a contact, so every dollar of marketing spend lost its trail at exactly the point the trail started to matter. Original source now persists onto the contact, the opportunity and the closed deal, which is what made cost per opportunity calculable for the first time in the company's life.

Weeks 7 to 11, stages and the cleanup nobody wanted to do

Stages were rewritten as exit criteria, meaning a condition another person can check rather than a feeling. The VP of Sales, not us, owned enforcement, which is the only version of this that survives contact with a quarter end.

We tested whether it had worked the cheap way. Before the rewrite, we gave two sales managers the same 25 live deals and asked them to stage each one independently. They agreed on 11. After, on a fresh 25, they agreed on 21. That test costs an hour and it is the only honest evidence that a stage definition is doing anything.

Then we resolved the stale pipeline, all 150 of it, with a loss reason from a picklist on every one.

That is when the win rate went from 34% to 22%.

It also did something to the sales cycle that is worth explaining, because it looks like bad news and is not. Reported average cycle length jumped from 74 days to well over a hundred that quarter, since deals that had been sitting open for up to 19 months finally resolved and landed in the average. It settled back to 71 the following quarter. Nothing about the sales process got slower. The measurement stopped ignoring its own worst cases.

Weeks 12 to 16, reporting and the job description

Reporting was rebuilt once, on the small set of fields that get filled reliably. The Monday pipeline report had been most of a day of someone's week. It became nobody's. We also started recording the forecast before the quarter rather than explaining it after, which nobody had been doing.

The nine tools became six. Two were doing the same job for different teams and one had not been logged into since the previous year. That saved about $1,400 a month, which is the least interesting part of it. What mattered was three fewer integrations to break at 2am and three fewer places for a contact record to disagree with itself.

Last, and this is the part they hired us for without knowing it, we wrote their RevOps job description. Not a template. Theirs, drawn from what four months of the work had actually demanded, with the one-time build separated from what recurs forever.

The system runs after the sale too

The three CS people in this story are the reason the engagement went to 16 weeks instead of 12.

Renewals were being booked as new business. That single mistake meant net revenue retention could not be calculated at all, and the retention number the board had been shown, comfortably north of 115%, was wrong in the company's favour. Churn risk lived in a CS manager's head and a private Slack channel.

We split expansion, contraction and renewal into their own event types and rebuilt the number. Actual NRR was 104%. Two quarters later it was 109%, and this time it was a measurement rather than a hope.

We also put a health flag on the account record, driven by three things the company already tracked and had never joined up: logins per licensed location over the trailing 30 days, open support tickets older than a week, and whether the original champion still worked there. The AE sees it before a renewal conversation rather than after.

None of that was in the original scope. It came out of the week-one audit and ranked fourth on the list, which is the argument for spending the first two weeks measuring.

The result

The honest headline is that the first thing we did to their numbers was make them worse.

Resolving 150 dead opportunities dropped the reported win rate from 34% to 22%. That was never a real decline. Over the next two quarters, with routing fixed and stages meaning something, it settled at 27% on a base of 60 opportunities a quarter, which sits comfortably inside the 20% to 35% range most B2B teams live in and below the 35% to 45% you would expect at their deal size. There is room left.

The same thing happened on the retention side. A number the board had been told was above 115% turned out to be 104% once renewals stopped being counted as new business. It was 109% two quarters later.

The 115% should have been the tell. Companies with contract values in the $25K to $50K range run a median net revenue retention around 102%, with the top quartile near 111%. A Series A company quietly outperforming the top quartile of its own ACV band is not usually a company with a great product. It is usually a company with a booking error. Both corrections went the same direction, and a founder who has sat through that twice tends to stop asking for the flattering version.

Median first touch went from about 31 hours to 9 minutes. Their first recorded forecast came in 11% over actual and the second came in 4% over, which is a good direction and two data points. We told them the same thing then that we would tell you now: come back after four quarters and we will know whether the forecast is accurate or whether it got lucky twice.

They made the hire about nine months in. They posted a RevOps Manager at $118K rather than the $165K director, because the build was finished and the job had become maintenance and iteration rather than excavation. That is $47K a year, and the engagement itself cost about what one quarter of the director hire would have, fully loaded.

The order matters more than the money. Hiring a RevOps lead into a broken system asks one person to run a rebuild, a rollout and a culture change at once, in their first 90 days, with no political capital and no map. Most of them leave inside a year, and then you are paying the search fee twice.

The reporting rebuild in this story is the same exercise as The Revenue Metrics Ladder. This team was reporting Level 3 metrics on Level 1 data, which is why nobody trusted the dashboard.

What they kept

Everything, throughout, not handed over at the end. Admin access from week one. A decision log with the reasoning, not just the change. Field and object documentation. The ranked failure list with what got fixed and what was consciously left. The routing rules in plain English next to the CRM configuration. The job description.

If we had disappeared in week 10, the team could have carried on. That is the test we design for, and it is the difference between a system you own and one you rent from your agency. We wrote up what that looks like in general in what you own when the agency leaves.

Frequently Asked Questions

Figures in this case study are representative of the engagement. Client details are anonymised to industry level.

← Back to all case studies